Re: Disable redo writing
From: Harel Safra <harel.safra_at_gmail.com>
Date: Thu, 27 Jan 2011 11:09:11 +0200
Message-ID: <AANLkTikrsbt8bQzdB-Nm18+zYoR6nGw8DFa8=D=tpOQs_at_mail.gmail.com>
>
> 1) How I can disable redo writing during the initial copy?
> I heard there is an hidden parameter for this.
> If you know about it and think that it is too dangerous to put on the
> list, please send it directly to me.
>
>
The parameter is called "_disable_logging". To quote Jonathan Lewis from a previous discossion in oracle-l titled "Disable logging in tablespace vs using hidden parameter _disable_logging": "Just as a side note, unless things have change in recent versions, _disable_logging still does all the work of generating redo log.
Date: Thu, 27 Jan 2011 11:09:11 +0200
Message-ID: <AANLkTikrsbt8bQzdB-Nm18+zYoR6nGw8DFa8=D=tpOQs_at_mail.gmail.com>
>
> 1) How I can disable redo writing during the initial copy?
> I heard there is an hidden parameter for this.
> If you know about it and think that it is too dangerous to put on the
> list, please send it directly to me.
>
>
The parameter is called "_disable_logging". To quote Jonathan Lewis from a previous discossion in oracle-l titled "Disable logging in tablespace vs using hidden parameter _disable_logging": "Just as a side note, unless things have change in recent versions, _disable_logging still does all the work of generating redo log.
The only change is that lgwr doesn't bother to write the buffer when called - basically bypassing the actual file-write call.
This means you still get redo latch activity, and checkpoints. Given the fact that lots of systems now have a workable buffer between the redo writer and the hardware, this means that _disable_logging may give you no appreciable benefit, unless you create MASSIVE redo log and avoid commits.
Regards
Jonathan Lewis"
Harel Safra
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Jan 27 2011 - 03:09:11 CST